Bug fix: treat old generations of reused objects as null

ChangeLog

@@ -1,3 +1,8 @@
+2024-01-07  Jay Berkenbilt  <ejb@ql.org>
+
+	* Bug fix: treat references to older generations of objects as
+	null.
+
 2024-01-06  Jay Berkenbilt  <ejb@ql.org>
 
 	* When recovering a file's xref table, attempt to find xref

include/qpdf/QPDF.hh

@@ -1038,6 +1038,7 @@ class QPDF
     QPDFObjectHandle makeIndirectFromQPDFObject(std::shared_ptr<QPDFObject> const& obj);
     bool isCached(QPDFObjGen const& og);
     bool isUnresolved(QPDFObjGen const& og);
+    void removeObject(QPDFObjGen const& og);
     void updateCache(
         QPDFObjGen const& og,
         std::shared_ptr<QPDFObject> const& object,

libqpdf/QPDF.cc

@@ -706,6 +706,19 @@ QPDF::read_xref(qpdf_offset_t xref_offset)
     // We no longer need the deleted_objects table, so go ahead and clear it out to make sure we
     // never depend on its being set.
     m->deleted_objects.clear();
+
+    // Make sure we keep only the highest generation for any object.
+    QPDFObjGen::set to_delete;
+    QPDFObjGen last_og;
+    for (auto const& og: m->xref_table) {
+        if (og.first.getObj() == last_og.getObj()) {
+            to_delete.emplace(last_og);
+        }
+        last_og = og.first;
+    }
+    for (auto const& og: to_delete) {
+        removeObject(og);
+    }
 }
 
 bool
@@ -1978,6 +1991,18 @@ QPDF::replaceObject(QPDFObjGen const& og, QPDFObjectHandle oh)
     updateCache(og, oh.getObj(), -1, -1);
 }
 
+void
+QPDF::removeObject(QPDFObjGen const& og)
+{
+    auto null = QPDFObjectHandle::newNull();
+    m->xref_table.erase(og);
+    if (isCached(og)) {
+        // Take care of any object handles that may be floating around.
+        replaceObject(og, null);
+    }
+    m->obj_cache.erase(og);
+}
+
 void
 QPDF::replaceReserved(QPDFObjectHandle reserved, QPDFObjectHandle replacement)
 {

qpdf/qtest/incremental.test

@@ -14,12 +14,8 @@ cleanup();
 
 my $td = new TestDriver('incremental');
 
-my $n_tests = 6;
+my $n_tests = 9;
 
-# Since the beginning but discovered at the time of releasing 11.8.0:
-# qpdf doesn't delete earlier generations of an object when they are
-# reused. See also EXPECT_FAILURE in object-stream.test and
-# linearization.test.
 $td->runtest("handle delete and reuse",
              {$td->COMMAND => "qpdf --qdf --static-id incremental-1.pdf a.pdf"},
              {$td->STRING => "", $td->EXIT_STATUS => 0},
@@ -27,11 +23,11 @@ $td->runtest("handle delete and reuse",
 $td->runtest("check output",
              {$td->FILE => "a.pdf"},
              {$td->FILE => "incremental-1-out.qdf"},
-             $td->NORMALIZE_NEWLINES | $td->EXPECT_FAILURE);
+             $td->NORMALIZE_NEWLINES);
 $td->runtest("check xref",
              {$td->COMMAND => "qpdf --show-xref incremental-1.pdf"},
              {$td->FILE => "incremental-1-xref.out", $td->EXIT_STATUS => 0},
-             $td->NORMALIZE_NEWLINES | $td->EXPECT_FAILURE);
+             $td->NORMALIZE_NEWLINES);
 $td->runtest("handle delete and reuse",
              {$td->COMMAND => "qpdf --qdf --static-id incremental-2.pdf a.pdf"},
              {$td->STRING => "", $td->EXIT_STATUS => 0},
@@ -40,11 +36,23 @@ $td->runtest("check output",
              {$td->FILE => "a.pdf"},
              # intentionally comparing incremental-2 with incremental-1-out
              {$td->FILE => "incremental-1-out.qdf"},
-             $td->NORMALIZE_NEWLINES | $td->EXPECT_FAILURE);
+             $td->NORMALIZE_NEWLINES);
 $td->runtest("check xref",
              {$td->COMMAND => "qpdf --show-xref incremental-1.pdf"},
              {$td->FILE => "incremental-2-xref.out", $td->EXIT_STATUS => 0},
-             $td->NORMALIZE_NEWLINES | $td->EXPECT_FAILURE);
+             $td->NORMALIZE_NEWLINES);
+$td->runtest("handle delete and don't reuse",
+             {$td->COMMAND => "qpdf --qdf --static-id incremental-3.pdf a.pdf"},
+             {$td->STRING => "", $td->EXIT_STATUS => 0},
+             $td->NORMALIZE_NEWLINES);
+$td->runtest("check output",
+             {$td->FILE => "a.pdf"},
+             {$td->FILE => "incremental-3-out.qdf"},
+             $td->NORMALIZE_NEWLINES);
+$td->runtest("check xref",
+             {$td->COMMAND => "qpdf --show-xref incremental-3.pdf"},
+             {$td->FILE => "incremental-3-xref.out", $td->EXIT_STATUS => 0},
+             $td->NORMALIZE_NEWLINES);
 
 cleanup();
 $td->report($n_tests);

qpdf/qtest/linearization.test

@@ -84,11 +84,6 @@ foreach my $base (@to_linearize)
 {
     foreach my $omode (qw(disable preserve generate))
     {
-        my $xflags = 0;
-        if ($base eq 'gen1')
-        {
-            $xflags = $td->EXPECT_FAILURE;
-        }
         my $oarg = "-object-streams=$omode";
         my $sdarg = "";
         if (($base eq 'lin-special') || ($base eq 'object-stream'))
@@ -100,13 +95,12 @@ foreach my $base (@to_linearize)
                      {$td->COMMAND =>
                           "qpdf -linearize $oarg $sdarg" .
                           " --static-id $base.pdf a.pdf"},
-                     {$td->STRING => "", $td->EXIT_STATUS => 0},
+                     {$td->STRING => "", $td->EXIT_STATUS => 0});
-                     $xflags);
         $td->runtest("check linearization",
                      {$td->COMMAND => "qpdf --check-linearization a.pdf"},
                      {$td->STRING => "a.pdf: no linearization errors\n",
                       $td->EXIT_STATUS => 0},
-                     $td->NORMALIZE_NEWLINES | $xflags);
+                     $td->NORMALIZE_NEWLINES);
         # Relinearizing twice should produce identical results.  We
         # have to do it twice because, if objects changed ordering
         # during the original linearization, the hint tables won't
@@ -117,17 +111,14 @@ foreach my $base (@to_linearize)
         $td->runtest("relinearize $base 1",
                      {$td->COMMAND =>
                           "qpdf -linearize $sdarg --static-id a.pdf b.pdf"},
-                     {$td->STRING => "", $td->EXIT_STATUS => 0},
+                     {$td->STRING => "", $td->EXIT_STATUS => 0});
-                     $xflags);
         $td->runtest("relinearize $base 2",
                      {$td->COMMAND =>
                           "qpdf -linearize $sdarg --static-id b.pdf c.pdf"},
-                     {$td->STRING => "", $td->EXIT_STATUS => 0},
+                     {$td->STRING => "", $td->EXIT_STATUS => 0});
-                     $xflags);
         $td->runtest("compare files ($omode)",
                      {$td->FILE => "b.pdf"},
-                     {$td->FILE => "c.pdf"},
+                     {$td->FILE => "c.pdf"});
-                     $xflags);
         if (($base eq 'lin-special') || ($base eq 'object-stream'))
         {
             $td->runtest("check $base ($omode)",

qpdf/qtest/object-stream.test

@@ -82,13 +82,9 @@ $td->runtest("generate object streams for gen > 0",
              {$td->COMMAND => "qpdf --qdf --static-id" .
                   " --object-streams=generate gen1.pdf a.pdf"},
              {$td->STRING => "", $td->EXIT_STATUS => 0});
-# qpdf 11.8.0 -- it was discovered that qpdf was incorrectly handling
-# references to older generations of reused objects in incrementally
-# updated files.
 $td->runtest("check file",
              {$td->FILE => "a.pdf"},
-             {$td->FILE => "gen1.qdf"},
+             {$td->FILE => "gen1.qdf"});
-             $td->EXPECT_FAILURE);
 
 $td->runtest("generate object streams for gen > 0",
              {$td->COMMAND => "qpdf --qdf --static-id" .

qpdf/qtest/qpdf/incremental-2-xref.out

@@ -1,7 +1,7 @@
 1/0: uncompressed; offset = 9
 2/0: uncompressed; offset = 63
 3/0: uncompressed; offset = 1069
-4/0: uncompressed; offset = 307
+4/1: uncompressed; offset = 948
 5/0: uncompressed; offset = 403
 6/0: uncompressed; offset = 438
 7/0: uncompressed; offset = 974

qpdf/qtest/qpdf/incremental-3-out.qdf

@@ -0,0 +1,79 @@
+%PDF-1.3
+%¿÷¢þ
+%QDF-1.0
+
+%% Original object ID: 1 0
+1 0 obj
+<<
+  /Pages 2 0 R
+  /Type /Catalog
+>>
+endobj
+
+%% Original object ID: 2 0
+2 0 obj
+<<
+  /Count 1
+  /Kids [
+    3 0 R
+  ]
+  /Type /Pages
+>>
+endobj
+
+%% Page 1
+%% Original object ID: 3 0
+3 0 obj
+<<
+  /MediaBox [
+    0
+    0
+    612
+    792
+  ]
+  /Parent 2 0 R
+  /Resources <<
+    /Font <<
+      /F1 4 0 R
+    >>
+    /ProcSet 5 0 R
+  >>
+  /Type /Page
+>>
+endobj
+
+%% Original object ID: 6 0
+4 0 obj
+<<
+  /BaseFont /Helvetica
+  /Encoding /WinAnsiEncoding
+  /Name /F1
+  /Subtype /Type1
+  /Type /Font
+>>
+endobj
+
+%% Original object ID: 5 0
+5 0 obj
+[
+  /PDF
+  /Text
+]
+endobj
+
+xref
+0 6
+0000000000 65535 f 
+0000000052 00000 n 
+0000000133 00000 n 
+0000000242 00000 n 
+0000000443 00000 n 
+0000000588 00000 n 
+trailer <<
+  /Root 1 0 R
+  /Size 6
+  /ID [<31415926535897932384626433832795><31415926535897932384626433832795>]
+>>
+startxref
+623
+%%EOF

qpdf/qtest/qpdf/incremental-3-xref.out

@@ -0,0 +1,5 @@
+1/0: uncompressed; offset = 9
+2/0: uncompressed; offset = 63
+3/0: uncompressed; offset = 135
+5/0: uncompressed; offset = 403
+6/0: uncompressed; offset = 438

qpdf/qtest/qpdf/incremental-3.pdf

@@ -0,0 +1,95 @@
+%PDF-1.3
+1 0 obj
+<<
+  /Type /Catalog
+  /Pages 2 0 R
+>>
+endobj
+
+2 0 obj
+<<
+  /Type /Pages
+  /Kids [
+    3 0 R
+  ]
+  /Count 1
+>>
+endobj
+
+3 0 obj
+<<
+  /Type /Page
+  /Parent 2 0 R
+  /MediaBox [0 0 612 792]
+  /Contents 4 0 R
+  /Resources <<
+    /ProcSet 5 0 R
+    /Font <<
+      /F1 6 0 R
+    >>
+  >>
+>>
+endobj
+
+4 0 obj
+<<
+  /Length 44
+>>
+stream
+BT
+  /F1 24 Tf
+  72 720 Td
+  (Potato) Tj
+ET
+endstream
+endobj
+
+5 0 obj
+[
+  /PDF
+  /Text
+]
+endobj
+
+6 0 obj
+<<
+  /Type /Font
+  /Subtype /Type1
+  /Name /F1
+  /BaseFont /Helvetica
+  /Encoding /WinAnsiEncoding
+>>
+endobj
+
+xref
+0 7
+0000000000 65535 f 
+0000000009 00000 n 
+0000000063 00000 n 
+0000000135 00000 n 
+0000000307 00000 n 
+0000000403 00000 n 
+0000000438 00000 n 
+trailer <<
+  /Size 7
+  /Root 1 0 R
+>>
+startxref
+556
+%%EOF
+
+% Delete object 4 and increment generation
+xref
+0 1
+0000000004 65535 f 
+4 1
+0000000000 00001 f 
+trailer <<
+  /Size 7
+  /Root 1 0 R
+  /Prev 556
+  /Gone 4 0 R
+>>
+startxref
+807
+%%EOF